A weak point of rule-based sentiment analysis systems is that the underlying sentiment lexicons are often not adapted to the domain of the text we want to analyze. We created a game-specific sentiment lexicon for video game Skyrim based on the E-ANEW word list and a dataset of Skyrim’s in-game documents. We calculated sentiment ratings for NPC dialogue using both our lexicon and E-ANEW and compared the resulting sentiment ratings to those of human raters. Both lexicons perform comparably well on our evaluation dialogues, but the game-specific extension performs slightly better on the dominance dimension for dialogue segments and the arousal dimension for full dialogues. To our knowledge, this is the first time that a sentiment analysis lexicon has been adapted to the video game domain.

In text generation, generating long stories is still a challenge. Coherence tends to decrease rapidly as the output length increases. Especially for generated stories, coherence of the narrative is an important quality aspect of the output text. In this paper we examine how narrative coherence is attained in the submissions of NaNoGenMo 2018, an online text generation event where participants are challenged to generate a 50,000 word novel. We list the main approaches that were used to generate coherent narratives and link them to scientific literature. Finally, we give recommendations on when to use which approach.